BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pyconde-pydata-2026//talk//GPV9SM
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-pyconde-pydata-2026-GPV9SM@pretalx.com
DTSTART;TZID=CET:20260415T173500
DTEND;TZID=CET:20260415T180500
DESCRIPTION:Most hyperparameter optimization (HPO) stops at the model bound
 ary. But what happens when your system relies on a complex chain of steps\
 , a short-horizon model\, a long-horizon model\, ensembles\, postprocesses
  etc? Tuning one piece in isolation often leads to sub-optimal global resu
 lts.\n\nIn this talk\, we explore how we used Ray to move beyond simple mo
 del tuning. We’ll dive into a "Pipeline-as-a-Trial" architecture where R
 ay acts as the brain\, triggering independent\, scalable cloud workflows (
  SageMaker Pipelines or Databricks Workflows) for every hyperparameter set
 .\n\n\nWe will discuss:\n* The architectural shift from tuning models to t
 uning pipelines\n* How to build the DAG/pipeline on Sagemaker/Databricks u
 sing declarative configs\n* How to use Ray to orchestrate heavyweight remo
 te jobs without bottlenecks.\n\nAttendees will learn how to optimize entir
 e pipelines (in a scalable manner on cloud) to minimize global metrics lik
 e WAPE\, rather than just local model loss.
DTSTAMP:20260523T180013Z
LOCATION:Dynamicum [Ground Floor]
SUMMARY:Holistic Optimization: Implementing "Pipeline-as-a-Trial" HPO with 
 Ray and Cloud Infra - Abdullah Taha
URL:https://pretalx.com/pyconde-pydata-2026/talk/GPV9SM/
END:VEVENT
END:VCALENDAR